Do AI vendors keep my data?
Claim Raven sends the text, files, images, and account context an AI feature needs to commercial AI providers. Depending on the feature and current settings, your private content can go to Anthropic, Google (Gemini), or OpenAI, and document pages can go to Microsoft Azure for text recognition when that option is turned on. xAI receives only public legal sources. Under their paid API terms, Anthropic, Google, OpenAI, and xAI do not use inputs and outputs for foundation-model training by default unless a customer opts in. Standard abuse-monitoring retention is commonly up to 30 days, but Google keeps Gemini API prompts and responses for 55 days, and any provider may keep limited content longer for safety-policy enforcement or legal obligations. Claim Raven applies best-effort masking only on supported text paths, so raw PDFs, images, and OCR can require sending an unredacted visual input, and text read from a file you attach to Ask Raven is sent without masking. It does not share account data with advertising vendors or data brokers.
Overview
Up to 30 days is not a universal deletion promise Provider terms differ by provider, endpoint, model, account setting, safety rule, legal obligation, and negotiated configuration. Google's Gemini API keeps prompts and responses for 55 days to detect policy violations. Anthropic can keep content flagged for a usage-policy violation for up to 2 years. Some provider features retain application state until it is deleted. Claim Raven does not promise that every request has zero data retention.
Which providers can receive your content
AI providers by feature. The provider used for a request depends on the feature and Claim Raven's current settings, and another listed provider can take over when the first one fails. Provider What it can receive Where Claim Raven uses it Anthropic (Claude) Page images and scanned PDF pages for text recognition, text read from your records for first-pass sorting, and short excerpts of your messages or files for small helper steps Reading scanned pages and images by default, first-pass sorting of Raven Scan medical-record notes, backup Raven Cipher page sorting, Raven Scribe chat, parts of Resume Raven, and small helper steps such as search-query cleanup, field extraction, and summary backups Google (Gemini) Ask Raven: your typed message, earlier messages, text read from attached files, and profile or claim context. Raven Eye: the original letter PDF or image and related profile context Ask Raven answers when Gemini routing is turned on, and Raven Eye letter and letter-packet analysis, including letters reviewed from your document library, unless that option is turned off OpenAI Your typed messages, earlier chat messages, text read from attached files, original PDFs and images, text drawn from your records and earlier tool results, and profile or claim context when AI Data Consent allows it Ask Raven backup answers and conversation summaries, Claim Workspace summaries and document reviews, Raven Eye letter analysis, backup analysis, and packet rollups, the DBQ Decoder, Raven Scan reports, synthesis, and claim briefs, evidence gap analysis, Raven Cipher page sorting, synthesis, and statements, Raven Nexus letter drafting and review, Pathfinder strategy synthesis, VA Bridge claim reads, Resume Raven, the statement tools, condition playbooks, and legal research tools, and short, lower-risk Ask Raven questions when that route is turned on Microsoft (Azure AI Document Intelligence) Page images and scanned PDF pages for text recognition Only for features where Claim Raven has turned on Azure text recognition, which is off by default; otherwise Anthropic reads those pages xAI Public BVA decisions and public regulatory or manual text only Public-source legal research. Your private documents and messages are not sent to xAI
What Claim Raven sends
Requested content: An AI feature may send the prompt, relevant chat context, extracted text, a file, image, screenshot, or scanned PDF page needed to perform the task you requested. Account context when authorized: Consent-dependent features may include relevant VA profile or claim context when AI Data Consent is enabled. The active Ask Raven conversation still needs its own messages and files to work even when cross-conversation or VA-profile personalization is off. Ask Raven routing: When Gemini routing is turned on, Gemini writes the first draft of an Ask Raven answer, including answers about attached files. An OpenAI model replaces that draft when Gemini fails or the draft fails Claim Raven's safety checks. A separate OpenAI route, when turned on, handles only short, lower-risk questions with no attachment and no detected identifier. Private-document routing: Most private-document analysis, including Raven Eye backup analysis, the DBQ Decoder, Raven Scan reports and synthesis, Resume Raven, and the statement tools, uses OpenAI. Raven Eye can send the original VA letter PDF or image to Google Gemini. Anthropic reads scanned pages and images by default and handles first-pass record sorting and small helper steps. Private documents and veteran messages go only to the providers listed above and are not sent to xAI. Public-source routing: Current xAI use is limited to app-initiated analysis of public BVA decisions and regulatory or manual text. Do not assume public legal-research processing and private veteran-document processing have the same provider or data rule.
What the provider terms do and do not mean
Training: Current paid API terms for Anthropic, Google, OpenAI, and xAI do not use API inputs and outputs to train foundation models by default unless the customer affirmatively opts in. Google's unpaid Gemini terms are different: Google may use that content to improve its products and people may review it. Microsoft's Azure Document Intelligence privacy page does not make a training statement. These are default commercial terms, not a promise that provider terms can never change. Safety and abuse monitoring: Anthropic, OpenAI, and xAI keep standard abuse-monitoring records for up to 30 days. Google keeps Gemini API prompts and responses for 55 days, and authorized Google staff can review flagged content. Azure Document Intelligence deletes submitted pages and results within 24 hours after analysis. A provider may retain limited content longer for safety-policy enforcement, legal obligations, or another documented exception. Endpoint, model, and account differences: Zero-data-retention or modified-monitoring programs require provider eligibility and configuration. Google says guaranteed zero data retention requires its separate Vertex AI service, which Claim Raven does not use for these features. Stateful endpoints, stored files, application objects, designated covered models, and negotiated terms can follow different rules. A request option such as store false is not proof that every provider copy or safety log is immediately deleted. Claim Raven use: Claim Raven does not use VA data, claim information, uploaded records, or AI content for advertising, marketing profiles, or sale. Contracted providers receive only the content needed for the requested service and remain subject to their applicable terms and Claim Raven instructions.
Masking has important limits
Claim Raven applies best-effort masking to common direct identifiers, such as Social Security numbers, VA file numbers, phone numbers, email addresses, street addresses, and labeled names, on supported text paths. Those include the message you type to Ask Raven, earlier chat messages, saved profile context, the personal statement tools, and C-File text in Raven Cipher. Masking does not cover everything. Text read from a file you attach to Ask Raven, text read from your files in Raven Eye, Raven Scan, and Claim Workspace, and the background summary of a long Ask Raven chat are sent without masking. Claim Raven also cannot reliably remove text or identifiers from an image or scanned PDF before a vision or OCR provider reads it, so a raw image, screenshot, or scanned page may reach the provider unredacted. Masking is not guaranteed to catch every identifier or medical detail.
Use the control that matches your goal
Share only what the task needs Remove unrelated pages and redact unnecessary identifiers before uploading when doing so will not undermine the task. Do not submit another person's records unless you have permission or other legal authority. Control future consent-dependent AI use Use AI Data Consent in Settings. Turning it off affects future consent-dependent access to VA profile, claim, cross-conversation, and cross-tool context; it does not erase the active chat's own messages or files and does not undo prior lawful processing. Delete a saved Claim Raven artifact Use the feature's control to remove a conversation, attachment, report, analysis, or other saved item. Deleting a Claim Raven record does not prove that a provider's safety log, legal hold, or backup copy was deleted at the same moment. Request broader deletion Use the verified account-deletion or consumer-health-data request process when appropriate. Claim Raven will notify relevant processors of a valid covered request, while provider and backup deletion can take longer where their terms or law permit. Check the current policy before sensitive use Provider lists, models, endpoints, routing, retention programs, and contractual controls can change. Review the current Privacy Policy and the notice shown by the feature before sending material you consider especially sensitive.
Current privacy and provider sources
Current privacy and provider sources Claim Raven Privacy Policy https://claimraven.com/privacy-policy Claim Raven Security and Privacy https://claimraven.com/security Anthropic Privacy Center: API data retention https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organization-s-data Anthropic Privacy Center: Covered Model retention practices https://privacy.claude.com/en/articles/15425996-data-retention-practices-for-covered-models Gemini API Additional Terms of Service: paid and unpaid services https://ai.google.dev/gemini-api/terms Gemini API usage policies: abuse monitoring and 55-day retention https://ai.google.dev/gemini-api/docs/usage-policies Gemini Developer API: zero data retention limits https://ai.google.dev/gemini-api/docs/zdr OpenAI API data controls and retention by endpoint https://developers.openai.com/api/docs/guides/your-data#default-usage-policies-by-endpoint xAI API security and data privacy FAQ https://docs.x.ai/developers/faq/security Azure AI Document Intelligence: data, privacy, and security https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/document-intelligence/data-privacy-security
Related privacy guidance
Related privacy guidance Review current AI providers and routing /privacy-policy#third-party-sharing Review AI Data Consent /privacy-policy#ai-data-consent See Claim Raven security controls /security Learn where Claim Raven stores data /help-center/privacy-and-security/faq-where-is-data-stored Keep files and support requests safer /help-center/privacy-and-security/feature-data-protection Learn how account deletion works /help-center/privacy-and-security/faq-delete-account
More information
Review the current AI-provider policy Open the Privacy Policy for the current provider list, routing boundaries, consent rules, and material retention qualifications. Open the Privacy Policy /privacy-policy#third-party-sharing